Seminar dan Call For Paper Munas Aptikom Politeknik
Telkom Bandung, 9 Oktober 2010
177
Tabel 3. Tingkat Risiko Stoneburner,2002
Dari pemetaan kerentangan dan kecenderungan terjadi maka dihasilkan level
risiko dapat diurutkan sebagai berikut tabel 4
Tabel 4. Level Risiko
Rating Jenis Risiko
score Rating
1 SDM 16.5 M
2 Sistem 10.7 M
3 Eksternal 5.3 L 4 Internal
5 L g.
Rekomendasi Kontrol Dari hasil penilaian maka direkomendasikan
beberapa kontrol yang dapat dilakukan terhadap risiko tabel 5
Tabel 5.Tabel Rekomendasi Kontrol Jenis Risiko
Rekomendasi kontrol SDM
• Training intensif
SDM •
Cross SDM •
Penambahan pegawai
• Outsourcing
Internal •
Analisis proses bisnis
• Desain proses
efektif Eksternal
• Maintenance
hardware •
Update antivirus •
Update aplikasi Sistem
• Perbaikan data
• Back up data
• Duplikat database
server •
Kontrol akses h.
Kontrol terpilih Dari hasil analisis biaya maka didapat kontrol
yang dipilih adalah sebagai berikut : 1.
Melakukan pelatihan SDM 2.
Melakukan penambahan pegawai 3.
Melakukan Update terhadap aplikasi 4.
Melakukan maintenace terhadap perangkat keras
i. Penanggung jawab kontrol terpilih
Untuk melakukan kontrol yang dipilih disusun panitia yang akan melaksanakan kegiatan
kontrol tersebut
.
Tabel 6 Penanggungjawab kontrol No
Kontrol Penanggungjawab
1 Melakukan pelatihan SDM
Bagian personalia dan Div TI
2 Penambahan Pegawai
Personalia 3 Melakukan
Update terhadap aplikasi
Divisi TI 4 Melakukan
maintenace HW Divisi TI
6. Kesimpulan dan Saran
Dari apa yang telah dibahas dapat disimpulkan bahwa :
1. Teridentifikasi berbagai risiko yang mungkin
menyerang perbankan, diantaranya adalah risiko operasional . Risiko yang terjadi pada
kegiatan operasional teknologi informasi perbankan ini melibatkan 4 jenis risiko, yang
diurutkan sesuai hasil penilaian level dimana risiko SDM 16,3 menempati urutan pertama
diikuti risiko sistem 10,3, eksternal 5,3 dan internal 5 secara berurutan.
2.
Dari apa yang dibahas dapat dilihat determinasi risiko dapat dilihat ternyata risiko
SDM menempati rating tertinggi, sementara sistem relatif kecil karena risiko baik dampak
maupun kecenderungan terjadinya sangat minim. Hal ini terbentuk dari tingkat
kecenderungan SDM untuk melakukan aktifitas yang terkait risiko dan dampak yang
dihasilkan mencapaim level menengah, sementara untuk level risiko lain untuk
kecenderungan risiko terjadi berada pada level
178
Seminar dan Call For Paper Munas Aptikom Politeknik Telkom
Bandung, 9 Oktober 2010 terendah, sedangkan untuk dampak yang
terjadi lebih ke dampak menengah dan beberapa poin tinggi, teteapi frekuensinya
tertutup oleh level kejadian yang relatif kecil
3. Masalah risiko yang terjadi terhadap
penggunaan teknologi informasi perbankan ditangani melalui pemetaan domain risiko
NIST SP 800 – 30 dengan memperhatikan beberapa faktor berikut :
a. Sistem
b. Jenis risiko operasional perbankan
c. Sumber risiko, dan
d. Level risiko
4. Dari hasil implementasi secara global
diketahui beberapa hal sebagai berikut :
a. Level risiko yang mengena pada Bank X
pada tingkat medium dan low. b.
Terdapat 4 rekomendasi yang akan diimplementasikan untuk minimalisir
risiko dari 7 rekomendasi yang diajukan. Saran untuk penelitian lebih lanjut dapat
dipetakan save implementation planning guard terhadap ke 4 empat rekomendasi yang dianjurkan
untuk kemudian dilakukan evaluasi terhadap efektifitas dari rekomendasi tersebut.
Pustaka [1] Bank
Indonesia .2003.
Pedoman Penerapan Risiko Bagi Bank umum, Bank
Indonesia.Jakarta [2]
Bank Indonesia .2007. Pedoman Penerapan Manajemen Risiko Teknologi Informasi di
Bank Umum. Bank Indonesia. Jakarta. [3] Basel Commite .2005. International
Convergence Of Capital Measurement and Capital Standar. Basel Comitte On Banking
Supervision. Basel. Switzerland. [4] Briean, O. John .2008. Management
Information system, Mc Graww – Hill. [5] Bouwer, W. Ryser, M. 2002. Risk
Management For Bank. ECOFIN Research and Consulting. Switzerland.
[6] Changqing G, Kezheng H. 2005. Comparison of Innovation Methodologies
and TRIZ. Shandong University. Jinan. P.R. China
[7] Darmini A,Putra I. 2010. Pemanfaatan Teknologi Informasi dan Pengaruhnya pada
Kinerja Individual BPR Kab. Tabanan. FE Universitas Udayana.
[8] Iman, E.
2008. http:www.erikiman.compublic01_definisi
_TI.pdf [9]
Froot K , Et. Al. 1994. A framework for Risk Management.
Massacusetts Institut
of Technology.
[10] Galorath D, 2006. Risk Magement Succes Factor. PM World Today. Vol.VIII, Issue
12. [11] Harjanto,Ludi.
http:www.wealthindonesia.comindex.php? option=com_contenttask=viewid=189
[12] Herarth, HSB. Herarth, T. 2007.Cyber Insurance : Copula Pricing Framework and
Implication for Risk Management..Departement of Acounting.St
Cathrine. Canada.
[13] ITGI.2009. Enterprise Risk : Identify, Govern and Manage IT Risk. IT Governance
Institute. USA. [14] Janakiraman, U. 2008. Operatinal Risk
Management Indian Bank In the Context of Basel II. Global Journal of Finance and
Banking Issues. Vo.2 No.2. [15] Kasmir.2008.
Bank dan Lembaga Keuangan Lainnya. Raja Grafindo Pt. Jakarta.
[16] Lubis, Muhammad E. 2008. Penetapan Model Bangkitan Pergerakan Untuk
Beberapa tipe Perumahan di Kota Pematangsiantar. Sekolah Pascasarjana.
USU. Medan.
[17] Nikolic , Boza. Dimitrizavic, L.2009. Risk Assesment of Information Technology
System. Issues In Informing Science and Technology. Vol.2009.
[18] Sutaryono,P.2003 .
http:avartara.comwaspadai-pemicu- internal-fraudmore-364
[19] Peter V, Peter R. 2006. Risk Management Model : an Empirical Assesment of The
Risk of Default. Euro Journal Publishing. [20]
Reppel, Milan. Tepley, Petr.2009. Operational Risk Analysis Scenario . ELBF
Seminar. Czech Republic. [21] Stela , M.I.2010. Evaluation of ICT on
Banking Effeciency Using Trancendental Lograthimi Production Fungciotn and
Camel Rating. International Journal Science and Technology. Vol.21.
[22] Stoneburner G, A. Goguen and A. Feringa, 2002.Risk Management Guide for
Information Technology System., Recommedation of National Institute of
Standards and Technology Special Publication 800-30.
[23] Wolingpirayat, J.2007.
E-payment Strategies of Bank Card Innovation. Journal of
Internet Banking And Commerce
Seminar dan Call For Paper Munas Aptikom Politeknik
Telkom Bandung, 9 Oktober 2010
179
Combined Fluctuation FeaturesFor Kid’s Song ClassificationBased on Mood Parameters Kadek Cahya Dewi
STMIK STIKOM BALI, Indonesia
cahya.amchygmail.com
Abstract Music is closely related to human psychology. A piece of music often associated with certain adjectives
such as happy, sad, romantic, and many more. The linkage between the music with a certain mood has been widely used in various occasions by people and music classification based on relevance to a particular
emotion is important. This research concerns in music classification system based on mood parameters with combined fluctuation features. The mood parameters used is based on Robert Thayers energy-stress
model which are exuberance happy, contentment relax, anxious and depression. All feature sets are based on fluctuation of modulation amplitudes in psychoacoustically transformed spectrum data, namely
the combination of rhythm patterns, rhythm histograms and statistical spectrum descriptors of the music. The system is tested using a set of song with various genre and the classification results are compared with
the mood obtained by child psychology experts. Clustering and classification method obtained by Self Organizing Map method.
Keywords: Self Organizing Map, music classification, mood classification, rhythm patterns, rhythm histograms, statistical spectrum descriptor
__________________________________________________________________________________________ 1. INTRODUCTION
Music is an art, entertainment and human activities that involve the voices of regular. Music
is all the possibilities that could happen to the voices sounds and silence to be organized into a
series of meaningful hearing. The meaning is not acquired verbal meaning, but the aural sense. The
meaning of aural harmony means the perceived sound when listening to music. For example when
listening a traditional song Yamko Rambe Yamko which originated from Irian Jaya, listeners may not
understand the intent of the song because they do not know the language that used in the lyrics or do
not understand whats played instruments. However, when hear it is very possible as a matter of making
nice to hear, making enthusiastic, sadness, grief or perhaps touching. Another example of the aural
sense is when listening to an instrumental song.
Music is closely related to human psychology. A piece of music often associated with
certain adjectives such as happy, sad, romantic, etc. The linkage between the music with a certain mood
has been widely used in various occasions by men. For example, in a musical film is used to reinforce
the atmosphere of the specific scene, the dramatic music used for background suspense scene, music
scene eager for war, the music is fun to use as a background scene of humor, etc. In addition to
support smart parenting programs will be very useful if the mothers know and understand the
classification of music based on mood parameters, so they can do song selection in accordance with
the desired conditions moods. For example when the children wake up just choose the happy song
full of spirit. There are many more examples that can not be mentioned here.
A number of researches on music classification based on mood have been conducted.
Feng et.al. classified music based on Dixon’s beat detection [3]. Li and Ogihara classified music using
Support Vector Machines SVM [7], whereas Yang and Lee using agent system [10]. Leman
et.al.based his classification on three level analysis: subjective judgments to manual-based musical
analysis to acoustical-based feature analysis[6]. Other researchers include Wang et.al. which uses
Support Vector Machines SVM [20], Wieczorkowska et.al. with K-Nearest Neigbhor
180
Seminar dan Call For Paper Munas Aptikom Politeknik Telkom
Bandung, 9 Oktober 2010 [21] and Baum with Naïve Bayes, Random Forest
and Support Vector Machines [1]. This paper will discuss the classification of
music based on relevance to a particular emotionmood. It will be compare the combined
fluctuation features. All feature sets are based on fluctuation of modulation amplitudes in psycho
acoustically transformed spectrum data, namely the combination of rhythm patterns, rhythm histograms
and statistical spectrum descriptors of the music. The system is tested using a set of song with
various genre and the classification results are compared with the mood obtained by child
psychology experts. Clustering and classification method obtained by Self Organizing Map method.
2. IMPLEMENTATION 2.1 Emotion Model
The emotion model used is the Robert Thayers energy-stress model [10] which consists of
exuberance happy, contentment relax, anxious anxiety and depression. Here is in Figure 1 is the
Thayer’s two dimensional model of emotion.
Figure 1: Thayer’s two-dimensional model of emotion
2.2 Feature Extraction
The feature extraction process is the process to get the pattern of a song. The following is the
feature extraction process in accordance with Rauber’s research [14] and the block diagram of
audio extraction is shown by Figure 2.
2.2.1. Preprocessing:
a. Audio decoding. Converting audio format into
Pulse Code Modulation PCM form. b.
Audio quality reduction. Audio quality is reduced from stereo to mono sound exchange.
Decreasing the number of channels from two to one and is done by taking the average of the two
channels. Audio sampling rate is changed from 44 KHz to 11 KHz.
c. Splitting music into segments with each segment
of size 6s in. Because of the time 6s deemed to have enough to get the impression from the style
of a piece of music.
2.2.2. Feature Extraction for Rhythm Patterns:
a. Transformation into a spectrogram, by first performing an FFT. The FFT window size used is
256 samples to meet the 23ms sampling of mp3 253 samples.
b. Groups frequencies into 24 critical frequency
bands to meet the Bark scale. Bark scale ranges from 1 to 24 Barks.
c. Calculating the spectral masking effects with spreading function.
d. Transform into decibel to form a base 10
logarithmic scale. e. Calculate the equal loudness level in Phon. Forty
Phon = 40 db-SPL tone at 1 kHz frequency. f. Calculating loudness sensation. One SONE = 1
kHz tone at 40 db-SPL. g.
Perform SONE transformation into rhythm patterns with the FFT.
h. Limit the amplitude modulation to 60. So that for every 24 critical bands 60 values is obtained for
modulation frequencies between 0 to 10Hz. This results in 1440 values representing the fluctuation
strength.
i. Filtering rhythm patterns using the Gaussian and the gradient.
2.2.3 Statistical Spectrum Descriptors
During feature extraction for the Rhythm Patterns, it was computed a Statistical Spectrum
Descriptor SSD for the 24 critical bands. From the SONE representation of the spectrum Sonogram,
we compute the following statistical moments for each critical band: mean, median, variance,
skewness, kurtosis, min-value and max-value, resulting in a 168-dimensional feature vector.
Exuberant, triumphant, carefree
Anxious, frantic, terror
Content, serene
Ominous, depression
stress energy
Seminar dan Call For Paper Munas Aptikom Politeknik
Telkom Bandung, 9 Oktober 2010
181
2.2.4 Rhythm Histograms
Contrary to the Rhythm Patterns and the SSD, this feature set does not contain information
per critical band. The magnitudes of each modulation frequency bin of all 24 critical bands
are summed up in order to form a histogram of modulation magnitude per modulation frequency.
This feature set contains 60 attributes, according to modulation frequencies between 0.168 and 10 Hz.
2.2.5 Combination of Feature sets
In order to evaluate the results on my past research entitle Kid’s Song Classification Based On
Mood Parameters Using Rhythm Patterns Features, now we try to use the combination of feature sets.
The different feature sets achieve largely different results depending on the database, i.e. the type of
music contained in the collection. As a consequence we are interested in the performance of combined
approaches, especially of the two sets with contrary results: SSD and Rhythm Histograms. The
combination is expected to represent a more generalized feature set with potentially better results
in a broader variety of musical styles. Moreover, we wanted to evaluate, whether classification without
the much higher-dimensional Rhythm Patterns feature set could achieve comparable results. The
following combinations of feature sets have been submitted to MIREX 2005 by Lidy and Rauber [8]:
– Rhythm Patterns + SSD 1608 dimensions
– SSD + Rhythm Histograms 228
dimensions –
Rhythm Patterns + SSD + Rhythm Histograms 1668 dimensions
Figure 2. Block Diagram Of Audio Feature Extraction